Mark Barnes / MarkVizion
Architecture Diagrams for AI-Native Systems
Workflow maps for agent loops, creative systems, and proof pipelines.
OnyxKraken Local Agent Loop
The key design principle is that the agent must observe state before acting and evaluate state after acting. That is what separates a workflow agent from a chatbot with tool calls.
- Screen / voice input
- Perception
- Memory + RAG
- Planner
- Tool router
- Desktop/API action
- Evaluation
- Updated memory
Creative Systems Engineering Loop
The creative layer is not decoration. It appears at the start as intent, in the middle as review, and at the end as a reusable product pattern.
- Creative intent
- Workflow map
- Prototype artifact
- Visual/system review
- Deployment
- User feedback
- Reusable pattern
AI Product Proof Pipeline
This pipeline turns output into authority. The app proves capability, the demo proves behavior, the case study proves reasoning, and the paper gives the concept a retrieval anchor.
- Problem wedge
- Capability match
- MVP build
- Demo / transcript
- Case study
- Paper
- Reusable framework